Hi Jonathan,
sorry, missed your mail completely.
Anyway, in 5.1 there will be tools to do all this automatically for you (also for more than 2 time points). Everything below is only for 2 time points.
Also about your questions: Something is wrong with you data, you should have the ?h.shpere.reg in the surf directories. Make sure your long runs finish without error. Make sure each subject has its own base (template).
You would usually divide by the time difference in any case if your time points are differently spaced between subjects: mris_calc -o new.mgh whatever.mgh div 1.2 (for example if time delta is 1.2 years)
average thickness is (for 2 time points): mris_calc -o avg.mgh tp1.mgh add tp2.mgh mris_calc -o avg.mgh avg.mgh div 2
and dividing by avg thickness mris_calc -o spc.mgh diff.mgh div avg.mgh
Note that spc (symmetrized percent change) is recommended over percent change with respect to time 1. Time 1 can be outlier and you get more power using the symmetrized version (with respect to the average).
Also note that mris_calc cannot deal with different input types (mgh and curv files),therefore use mri_convert to convert a curv file into mgh format, e.g. mri_convert lh.thickness lh.thick1.mgh
For more than 2 time points you could fit a GLM within each subject to compute the average and slope per time unit. But it depends on what you are trying to achieve, one could also compare rate of change in first year vs. rate of change in second year for a 3 tp study (each a year apart). Or more complex longitudinal models....
Best, Martin
On Wed, 2011-03-09 at 14:44 -0500, Jonathan DuBois wrote:
Hi Freesurfer users,
I have been working from the list posting below for vertex comparison of longitudinal thickness data, although there are a couple issues I was hoping someone could help with.
Using FS v5.0 with the fixed binaries
, I have put together the following steps for data comparison following the third step of the longitudinal stream with their associated commands and questions (highlighted in bold). I am sorry if it is a lot of information, although I thought it might be useful for other users working with longitudinal analysis.
Thanks
Jonathan
Martin Reuter mreuter at nmr.mgh.harvard.edu
Tue Jun 29 16:23:20 EDT 2010
Hi Sandra,
(and the list, I think this might be interesting for others)
- the longitudinal ?h.thickness files are in registration (within
subject)
so in order to compute the difference or the rate you can simply use:
mris_calc -o lh.outdiff sdir2/surf/lh.thickness sdir1/surf/lh.thickness
and for the rate, you can divide by time 1 thickness or by the average
thickness. If the time points are not equally spaced across subjects,
you can also divide by difference in time.
It is best to smooth the thickness files before you do all this (i.e.
within time point smooth the ?h.thickness files with mri_surf2surf).
- then map everything to your template (e.g. fsaverage) with
mri_surf2surf
- there you can stack the thick-differences or thick-rates with
mri_concat
- do the glm
Best
Gruesse, Martin
- Smooth surface data:
mri_surf2surf --s subj1_tp1.long.template --sval subj1_tp1.long.template/surf/lh.thickness --tval subj1_tp1.long.template/surf/lh.thickness.fwhm10.mgh --fwhm-src 10 --hemi lh
mri_surf2surf: could not read surface /space/mdeh2/1/halgdev/emcbkp/1/halgdev/projects/nyuproj/subjects/subj1_tp1.long.template/surf/lh.sphere.reg
No such file or directory
--The longitudinal processing does not create a ?h.sphere.reg or a ?h.sphere for running mri_register. While, the registration is specific to the longitudinal analysis, could I simply copy over the ?h.sphere.reg from the original subject, since it is the same brain? If not, how should this be generated?
- Compare time point differences within subjects using mris_calc and
one of the three options:
a. divide tp2 by tp1 - [mris_calc -o lh.NC024_thick_diff_fwhm10.mgh subj1_tp2.long.template/surf/lh.thickness.fwhm10.mgh div subj1_tp1.long.template/surf/lh.thickness.fwhm10.mgh]
b. divide by average thickness
c. divide by between scan time difference
--Option (a) is straight forward, although option (c) would be ideal given that our subjects vary in time between scans. How would option (b) or © be set up with mris_calc? Would mris_calc work for more than two time points?
- Register subjects' between scan difference map to fsaverage
mri_surf2surf --s lh.NC024_thickness_diff_fwhm10.mgh --trgsubject fsaverage --tval lh.NC024_thickness_diff_fsaverage.fwhm10.mgh --fwhm-src 10 --hemi lh
--I haven't gotten to this step yet, but I'm not sure that it will work given that there is only a .mgh file without an associated subject? Also, similar problem to question one (no sphere.reg).
Stack subjects' difference maps using mri_concat
Run GLM using mri_glmfit
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer